João Amador** | Ana Cristina Soares**
Abstract
This article estimates price-cost margins for the Portuguese markets in a context of
imperfect competition in the labour market. The database used includes virtually the
universe of Portuguese firms for the period 2005-2009. The results strongly reject
the hypothesis of perfect competition in both labour and product markets. Estimated
price-cost margins are very heterogeneous across markets and the average for the
overall economy ranges between 25 and 28 per cent, depending on the variables used
to weight each market. In addition, the tradable sector presents a lower price-cost
margin than the non-tradable sector. According to the methodology used, workers’
bargaining power in the Portuguese economy is approximately 13 per cent, without
a clear distinction between tradable and non-tradable sectors. Finally, workers’
bargaining power is positively correlated with price-cost margins.
1. Introduction
Competition in the product market is a key ingredient for an efficient allocation of resources in the
economy, thereby promoting a higher aggregate welfare. Therefore, the identification of markets where
there are large deviations from the perfect competition paradigm is an important policy concern. From
a theoretical point of view, market power relates to firms’ ability to increase profits by sustaining prices
above marginal costs. However, establishing robust measures of competition is a strong challenge both
from a theoretical and empirical point of view.
This article uses the methodology presented by Roeger (1995), which closely relates to the approach
proposed by Hall (1988), to test whether there is a significant gap between prices and marginal costs
in Portuguese markets, i.e., how distant are markets from the perfect competition paradigm. The
methodology proposed by Hall (1988) for the estimation of price-cost margins is based on the relation
between the Solow residual and the growth rate of inputs. However, this relation cannot be estimated
by standard econometric methods such as OLS, since input growth rates are likely to be correlated with
technological progress, which is not observable. In this context, Hall (1988) suggests the use of instrumental variables. However, finding suitable instruments is, in general, a severe obstacle. More recently,
other authors propose the use of the generalized method of moments, such as Dobbelaere (2004), or
the use of a control function, as Olley and Pakes (1996) and Levinsohn (1993).
An alternative methodology was proposed by Roeger (1995). This methodology uses the difference
*
The authors thank António Antunes, Nuno Alves, Mário Centeno, Jorge Correia da Cunha, Ana Cristina Leal,
José António Machado and Pedro Portugal for their comments and Lucena Vieira for her support in questions
related to the database. The opinions expressed in the article are those of the authors and do not necessarily
coincide with those of Banco de Portugal or the Eurosystem. Any errors and omissions are the sole responsibility of the authors.
** Banco de Portugal, Economics and Research Department.
III
75
Articles
COMPETITION IN THE PORTUGUESE ECONOMY:
ESTIMATED PRICE-COST MARGINS UNDER
IMPERFECT LABOUR MARKETS*
between the Solow residuals obtained through profit maximization and cost minimization problems of
the firm, as a way to overcome the main source of endogeneity in the formulation of Hall (1988). In the
standard version of these methodologies, constant returns to scale and the existence of homogeneous
III
BANCO DE PORTUGAL | ECONOMIC BULLETIN • Spring 2013
76
inputs that adjust instantly in perfectly competitive markets are generally assumed. However, the literature
has discussed the validity of these assumptions, particularly with respect to perfect competition in the
labour market. In fact, recent empirical evidence suggests that the level of product market imperfection
is significantly underestimated when the degree of imperfection in the labour market is ignored.
In this context, both methodologies were modified to estimate simultaneously product and labour market
imperfections, measured by the price-cost margin and workers’ bargaining power, respectively. Beyond
the explicit test of perfect competition, one of the advantages of both Hall (1988) and Roeger (1995)
methodologies is that differences between technologies across sectors are partially taken into account
by the use of production functions.
This article contributes to the assessment of competition in the Portuguese economy, complementing
the alternative approaches presented in Amador and Soares (2012a,b). A distinctive feature of the article
is the coverage of a large number of markets in the economy (including services) and the distinction
between tradable and non-tradable sectors. This distinction is relevant given the potential disciplinary
effect of international competition and the nature of the sectoral adjustment process currently underway
in the Portuguese economy. Other distinctive features are the use of firm-specific measures of the user
cost of capital and depreciation rates, the inclusion of tangible and intangible assets, and the test for
sample selection bias.1 The data used in this article is based on information on the annual accounts of
Portuguese firms reported under Informação Empresarial Simplificada (IES) for 2005-2009.
The article concludes that the assumption of perfect competition in Portuguese product markets is widely
rejected, though there is substantial heterogeneity in price-cost margin estimates. Allowing for imperfect
competition in the labour market, the estimated price-cost margin for the overall economy ranges between
25 and 28 per cent, depending on the variables used to weight each market. Additionally, the price-cost
margin in the tradable sector is lower than in the non-tradable sector. Similarly, perfect competition in
the labour market is rejected in around 75 per cent of the markets. The workers’ average bargaining
power in the Portuguese economy lies between 12 and 14 per cent, according to weights considered
for each market, without a clear distinction between tradable and non-tradable sectors. Nevertheless,
there is a significant dispersion across markets. Consistent with the results presented in the empirical
literature, estimates for workers´ bargaining power are positive and strongly correlated with price-cost
margins across markets in the Portuguese economy.
The article is organized as follows. The next section briefly describes the methodology used in the estimation of price-cost margins under competitive and imperfect labour markets. Next, section 3 describes
the database and presents the definition of the variables. Section 4 presents the results, highlighting the
difference between tradable and non-tradable sectors. Section 5 presents some concluding remarks.
2. Methodology
Technological progress and market power are strongly related from a theoretical and empirical point of
view. The seminal contribution of Solow (1957) introduced growth accounting to identify the role of
technological progress. Later, Hall (1988) and Roeger (1995) relaxed the assumption of perfect competition
in the product market, allowing for the estimation of markups. The standard formulation relies on the
assumptions of efficient and homogeneous input markets, instantaneous adjustment of all input factors
and constant returns to scale. Subsequently, the assumption of perfect competition in the labour market
was relaxed, allowing for the joint estimation of price-cost margins and workers’ bargaining power.
1 For more details on the methodology used in this article and additional results see Amador and Soares (2013).
2.1 Price-cost margin estimation under competitive labour markets
Considering a neoclassical production function, the assumption of efficient input markets drives the
standard equality between the value of marginal productivity and the corresponding price of the input.
III
Consequently, input elasticities correspond to their weight in output. Therefore, in the presence of market

1
1
SR   1   (q  k )  




(1)
where  is the markup,  represents the growth rate of Hicks-neutral technological progress and q and
k are the logarithms of output and capital, respectively. Therefore, the classical price-cost margin can


be obtained from the estimate of the parameter 1  1 /  in equation 1. This parameter corresponds
to the Lerner index, defined as (P  MgC ) / P , where P and MgC represent the price and marginal
cost, respectively. However, the last term in equation 1 is not observable, thus the OLS estimator is
inconsistent. The solution proposed by Hall (1988) consists in using instrumental variables. However, it
is usually difficult to obtain suitable instruments and results tend to be sensitive to this choice. In this
context, Roeger (1995) proposed an alternative approach.
Considering the firm´s dual problem, i.e., cost minimization for a given level of output, along with the
assumption of imperfect competition in the product market and constant returns to scale, the Solow
residual of the dual problem (SRd) is:
SRd  (1 
1

)(p  r ) 
1


(2)
where p is the logarithm of the price and r is the logarithm of the cost of capital. Finally, adding the
primal and dual Solow residuals (equations 1 and 2), it is possible to write:

1
SR  SRd   1   (p  q )  (r  k )


(3)
Consequently, the term related to technological progress in equation 3 is eliminated, solving the inconsistency problem mentioned above. This approach allows for the estimation of the price-cost margin
consistently by OLS. Furthermore, this formulation avoids the use of deflators, which is a source of
measurement error, particularly when firm level data is used. However, a measure of the cost of capital
is required.
2.2 Price-cost margin estimation under imperfect labour markets
In the previous subsection market power was estimated assuming that workers receive perfectly competitive wages, i.e., assuming that their bargaining power is null. However, this assumption is not supported
by empirical evidence.
The approaches suggested by Hall (1988) and Roeger (1995) can be modified to account for imperfect
competition in the labour market (see Crépon et al., (2005), Dobbelaere (2004) and Abraham et al.,
(2009)). Under imperfect labour markets, wages (W) and the number of workers (L) are simultaneously
chosen according to a standard Nash bargaining problem, which involves sharing the surplus between
firms that maximize profits and workers whose utility depends on employment and wages, that is:

max   (W  W )L  (PQ  WL)(1 )


L,W
(4)
Articles
77
power and assuming constant returns to scale, the Solow residual (SR) can be rewritten as follows:
where W is the reservation wage (related to the best alternative wage in the labour market and unem-
III
BANCO DE PORTUGAL | ECONOMIC BULLETIN • Spring 2013
78
ployment benefits), P and Q represent the price and quantity sold, respectively. In addition, 1    0
represents the bargaining power of the workers, where   0 corresponds to competitive labour markets
and   1 to a total appropriation of the firm’s surplus by the workers. 2 In this context, assuming imperfect competition and an isoelastic demand function, the Solow residual can be written as:

   L
1
1
SR   1   (q  k )  
 (  1)  l  k )  

1







(5)
where  L represents the weight of labour costs in output. The dual counterpart of this problem is:

   L
1
1
SRd   1   (p  r )  
 (  1)  w  r )  



1 
(6)
where w is the logarithm of wages. Thus, allowing for imperfect competition in the labour market and
assuming constant returns to scale, the modified Roeger (1995) approach is:

1

( L  1)[(l  w )  (r  k )]
SR  SRd   1   [(p  q )  (r  k )] 

(1   )

(7)
This equation allows for the joint estimation of price-cost margins and workers’ bargaining power. The
exclusion of the last term induces a bias in the price-cost margin estimate, which is higher the higher
the bargaining power, the weight of labour costs in output and the larger the difference between the
growth rate of nominal labour and capital costs.
3. Database and variable definitions
3.1. Database description
The data used in this article draws on the annual accounts of Portuguese firms reported under Informação Empresarial Simplificada (IES) for 2005-2009.3 This database provides very detailed information
on items of the balance sheet and income statements for virtually the universe of non-financial firms.
The initial raw dataset coincides with the one used in Amador and Soares (2012a,b). However, at odds
with these articles, the information drawn from Central de Balanços for 2000-2004 was not considered.
Since Central de Balanços contains information on a sample of Portuguese firms, comprising mainly large
ones, the final set of information was insufficient to ensure the significance of the estimated parameters.
On the contrary, in the case of IES, despite being available on a comparable basis for a limited number
of years, its almost universal coverage provides a substantial set of observations.
2 There are alternative models of negotiation between firms and workers where wages and number of workers
are decided sequentially (see, e.g., Walque et al., (2009)). In addition, there are methodological options in the
Nash bargaining setup that may change results, namely the firm´s thread point at the moment of negotiation. In
this context, the definition of capital stock (gross or net), as well as the use of GVA alternatively to output can
also change results.
3 Although IES formally began in 2006, it included a report for 2005. For this reason, for the purpose of this
article, IES is considered from 2005 onwards.
Some observations were eliminated from the database to ensure robust estimations. Firstly, firms reporting
less than two consecutive observations were eliminated. Additionally, only firms reporting strictly positive
sales, labour costs, intermediate inputs and net capital stock (tangible and intangible) were considered.
in total sales outside the [0,1] range were excluded. Moreover, observations below the 1st percentile
and above the 99th percentile in the distribution of growth rates of sales, labour costs, intermediate
inputs and tangible and intangible assets were excluded. Thirdly, consistent with profit maximization in
the long run, firms exhibiting negative operational profits were withdrawn, representing approximately
22 per cent of the observations in the database. However, this option may increase the potential for
the existence of a sample selection bias. Although this problem is typically disregarded in the literature,
in this article the impact of selection bias is assessed through the two-step Heckman (1979) procedure.
Finally, sectors as “Agriculture, Mining and Quarrying”, “Education” and “Health” were disregarded
given their low share in total gross value added (GVA) or the significant relevance of the general government in the functioning of the market.
Given the reduced number of observations for each firm over the period considered, price-cost margins
were estimated at market level, i.e., we assume that price-cost margins and bargaining power are the
same for all firms within each market. Nevertheless, it is necessary to establish a criterion to define
markets. In order, to overcome the well known difficulties in establishing relevant markets, the standard
approach in the literature is to use an economic activity classification. Similarly to Amador and Soares
(2012a,b), markets are defined at 3-digit level in NACE Rev. 1. However, markets with less than 5 observations for a given year were eliminated. Overall, the article considers a total of 156 markets, 108 of
which are considered tradable and 48 as non-tradable. As discussed in Amador and Soares (2012a), the
set of tradable markets includes all manufacturing markets plus those where the exports to sales ratio
exceeds 15 per cent.4 In this sample, the non-tradable sector represents 56 per cent of GVA, 61 per cent
of sales and 54 per cent of total employment in the period 2006-2009.
3.2 Definition of variables and descriptive statistics
The set of variables required to estimate equation 7 is relatively large. Firstly, output corresponds to sales
of goods and services, and its growth rate is pt  qt . Secondly, labour costs are given by nominal
wages and other benefits including social security contributions, and its growth rate is represented by
lt  wt . Thirdly, shares of labour costs and intermediate inputs (  L and  M ) consist of the ratios
of labour costs and costs of goods and services to sales, respectively. Chart 1 displays the distribution
of these shares for Portuguese firms in 2008, distinguishing between those operating in tradable and
non-tradable sectors. The average share of labour costs and intermediate inputs are 25 and 62 per cent,
respectively. The distribution of labour cost shares is positively skewed, presenting greater dispersion in
the tradable sector. In contrast, the distribution of intermediate inputs shares is negatively skewed in the
non-tradable sector and closer to a Gaussian distribution in the tradable sector.
The estimation of equation 7 also requires information on the stock of capital and its user cost. Differently
from most studies, the stock of capital considered in this article includes both tangibles and intangibles.
If intangibles are dismissed, results can be substantially biased, particularly in services markets where
these assets tend to have an extremely relevant role.
The user cost of capital is the price to pay for hiring or purchasing one unit of capital services and includes
a measure of the financial cost of capital and the depreciation rate. Unlike most studies in the literature,
this cost was calculated at firm level, which is likely to reduce measurement error. Following Hall and
4 Note that the set of markets considered in the article does not fully coincide with the one used in Amador and
Soares (2012a,b) basically due to the exclusion of the information obtained under Central de Balanços (20002004).
III
79
Articles
Secondly, observations associated to depreciation rates and share of labour costs and intermediate inputs
Chart 1
DISTRIBUTION OF LABOUR AND INTERMEDIATE INPUT SHARES ON SALES AT FIRM-LEVEL IN 2008
III
a) Labour share
b) Intermediate input share
3
Density
2
3
1
Density
2
1
0
0
BANCO DE PORTUGAL | ECONOMIC BULLETIN • Spring 2013
4
4
80
0
.2
Tradable
.4
.6
Non-tradable
.8
1
Overall economy
0
.2
Tradable
.4
.6
Non-tradable
.8
1
Overall economy
Source: Author’s calculations.
Jorgenson (1967), the user cost of capital of firm i in year t is defined as ri,t  (ii, t  PˆI , t  i, t )PI , t , where
i is the financial cost of capital,  is the depreciation rate, PI , t and PˆI , t represent the level and
i, t
i, t
growth rate of investment goods prices, respectively. These elements derive from the standard equation
that relates the value of an asset to the discounted real flows of rentals expected over its lifetime.5
The depreciation rate at firm level is calculated as the ratio of total depreciations in year t to gross capital
stock in year t-1, i.e., for firm i in year t, i, t  depreciationi, t / K i, t 1 . The calculation of firm-level
depreciation rates makes it possible to capture some of the heterogeneity in the stock of capital. Chart
2 a) presents the distribution of the depreciation rate for Portuguese firms in 2008. The distribution is
positively skewed and the average for the overall economy lays around 10 per cent, with no significant
differences between firms in tradable and non-tradable markets. These figures are in line with those
used in similar articles. For example, Christopoulou and Vermeulen (2012) use a rate of 8 per cent with
longitudinal data, Boulhol et al., (2006) uses rates of 5 and 7 per cent, while Konings and Vandenbussche
(2005) assume a depreciation rate of 10 per cent.
While the calculation of the depreciation rate is relatively straightforward, the calculation of financial
cost of capital is more complex. This article assumes that the financial cost of capital is given by the ratio
between interest and financial debt for each firm and year. Thus, it is assumed that funding through equity
is equivalent to funding through debt. Chart 2 b) shows the distribution of the financial cost of capital
of Portuguese firms in 2008. The distribution is positively skewed, with an average of approximately 15
per cent and a median of 10 per cent. Additionally, the density in the tail that corresponds to lower costs
of capital is higher in the non-tradable sector than in the tradable sector. Finally, regarding the deflator
of investment goods ( PI , t ), it was obtained directly through national accounts.
5 For more details on the methodologies used to measure the capital stock and its user cost see OECD (2001).
Chart 2
DISTRIBUTION OF DEPRECIATION RATE AND FINANCIAL COST OF CAPITAL AT FIRM-LEVEL IN 2008
a) Depreciation rate
III
b) Financial cost of capital
Density
4
4
0
0
2
2
Density
6
6
8
Articles
8
81
0
.1
Tradable
.2
.3
Non-tradable
.4
0
.1
.2
Tradable
Overall economy
Non-tradable
.3
Overall economy
Source: Author’s calculations.
In order to avoid a substantial loss of observations, the financial cost of capital of the firms that report
no debt, interest payments or ratios outside the [0,1] range was considered equal to the average of the
respective market in each year. Chart 3 displays the distribution of the user cost of capital of Portuguese
firms, using the imputation referred above. This distribution is positively skewed with an average of
about 20 per cent.
Chart 3
0
2
Density
4
6
DISTRIBUTION OF THE USER COST OF CAPITAL AT FIRM-LEVEL IN 2008
0
.1
Tradable
.2
.3
Non-tradable
.4
.5
Overall economy
Source: Author’s calculations.
Note: The distribution displayed in the chart corresponds to the real financial cost of capital added to the depreciation rate.
4. Results
In this section we test the paradigm of perfect competition in Portuguese product markets in the period
III
2006-2009, allowing for imperfect labour markets, i.e., estimating equation 7 for each market and
clustered errors (benchmark specification). In addition, regressions with fixed effects, random effects and
BANCO DE PORTUGAL | ECONOMIC BULLETIN • Spring 2013
distinguishing those with a tradable and non-tradable nature. The equation is estimated by OLS with
82
the two-step Heckman procedure are also estimated to ensure robust results. Furthermore, aggregations
for some sectors are presented, as well as for the overall economy.
The perfect competition paradigm is widely rejected in Portuguese product markets. At a significance
level of 5 per cent, estimated price-cost margins are statistically different from zero for virtually all
markets considered (95 per cent of the markets). Chart 4 a) ranks estimated price-cost margins from the
highest to the lowest, uncovering a substantial heterogeneity across markets. Price-cost margins range
between a minimum of 6 per cent and a maximum of 62 per cent. The comparison between tradable
and non-tradable sectors suggests lower competition intensity in the latter, with unweighted price-cost
margins of 26 and 29 per cent, respectively. This difference is slightly higher when manufacturing and
non-manufacturing sectors are compared. The price-cost margin for the overall Portuguese economy
stands at 27 per cent.
Given the relevance of the results in terms of policy, the comparison of price-cost margins obtained
through different econometric approaches is particularly important. Chart 4 b) reports price-cost margins
estimated by fixed effects, random effects and two-step Heckman procedure for each market, sorted
according to the benchmark specification.6 It should be noted that the rank of markets obtained through
the different specifications is largely unchanged, implying that the identification of markets with a
Chart 4
PRICE-COST MARGIN FOR EACH MARKET IN THE PERIOD 2006-2009
a) Benchmark specification
65
60
55
b) Alternative specifications
65
Tradable
Non-tradable
Average (Tradable)
Average (Non-tradable)
60
55
50
45
45
40
40
35
Per cent
Per cent
50
35
30
30
25
25
20
20
15
15
10
10
5
5
0
0
OLS
Fixed effects
Random effects
Two step-Heckman
Source: Author’s calculations.
Note: Each market corresponds to a 3 digit level in NACE Rev. 1 classification. Black bars identify non-tradable markets as defined
in Amador and Soares (2012a). Coefficients were obtained through OLS regressions with cluster errors, for each market (benchmark
specification). Grey bars correspond to coefficients not significant at 5 per cent, in at least one specification.
6 The two-step Heckman procedure was used to test and correct the potential sample selection bias associated
with the exclusion of a substantial number of firms with negative operational profits. The inverse Mills ratio is
significant for around 30 percent of the markets, at a 5 per cent significance level. The explanatory variables in
the participation equation are firm´s age, sales and lagged total assets, in logarithm. Furthermore, the introduction of annual dummies in the remaining econometric approaches did not affect the results, thus they were not
included. The Hausman test was also performed for each market and random effects were rejected in around
45 per cent the markets at a 5 per cent significance level.
potentially less intense competitive environment does not change. The share of markets where there is
statistical evidence to reject the perfect competition paradigm is below 8 per cent for all specifications,
and these markets belong exclusively to the manufacturing sector.
dology allows for the existence of imperfect competition in the labour market, i.e., when workers hold
some bargaining power. Under this assumption, the regression captures the overall surplus extracted by
the firm to the consumer through its market power, including the part that is transferred to the workers
through their bargaining power in the labour market. In fact, by assuming perfect competition in the
labour market (null bargaining power for the workers), labour costs are incorrectly assumed to translate workers’productivity, thus underestimating firm´s market power. Chart 5 illustrates this result by
comparing price-cost margins presented above with those obtained assuming perfect competition in the
labour market. The average underestimation is 11 p.p., though in some markets the bias reaches values
above 35 p.p.. The results in the empirical literature have also pointed a substantial underestimation.
Bassanetti et. al. (2012) refers an underestimation of 10 p.p.. Considering only the manufacturing sector,
Dobbelaere (2004) reports a higher underestimation, around 20 p.p.. Still, there is a high correlation
between estimated price-cost margins in both contexts (80 per cent), i.e., markets associated to lowest
competition intensity do not change substantially.
The estimate for the term  / (1   ) in equation 7 makes it possible to recover the parameter for the
workers’ bargaining power ( ) in each market. Chart 6 a) reports workers’ bargaining power in each
market sorted in descending order. As reported for the product market, the assumption of perfect
competition in the labour market is widely rejected (in about 75 per cent of the markets, at a significance level of 5 per cent). This percentage is higher in the non-tradable (85 per cent) than in tradable
sector (72 per cent).
Chart 5
PRICE-COST MARGIN UNDER COMPETITIVE AND IMPERFECT LABOUR MARKETS | PER CENT
Price-cost margin under imperfect labour markets
65
60
55
50
45
40
35
30
25
20
15
Tradable
Non-tradable
10
5
0
0
5
10
15
20
25
30
35
40
45
50
55
60
65
Price-cost margin under perfect labour markets
Source: Author’s calculations.
Notes: Each market corresponds to a 3 digit level in NACE Rev. 1 classification. Black dots identify non-tradable markets as defined
in Amador and Soares (2012a). Coefficients were obtained through OLS regressions with cluster errors, for each market.
III
83
Articles
One of the results in the literature is that estimates for the price-cost margin are higher if the metho-
Workers’ bargaining power is very heterogeneous, reaching values higher than 30 per cent in specific
markets of “Transports” and “Real estate activities” but also very low figures in markets related to
“Trade” and the manufacturing sector. Negative values are abnormal and are associated to non signithe labour market. Unweighted average bargaining power for the overall economy stands at 14 per
cent, close to the figures found for tradable and non-tradable sectors. Regarding the results for different
formulations, chart 6 b) overlaps estimates sorted according to the benchmark specification. The results
are broadly consistent, though it can be seen that some estimates obtained using fixed effects differ
from the benchmark specification but the overall rank is maintained.
As it is suggested in the empirical literature, results show that the degree of imperfection in the product
market is closely related to the degree of imperfection in the labour market. The correlation between
price-cost margins and workers’ bargaining power across markets is around 81 per cent (Chart 7). For
example, Estrada (2009) reports a correlation of 50 per cent for several EU countries for the period
1980-2004. Considering only the manufacturing sector, Boulhol et al., (2006) studied 20 markets in the
UK in the period 1988-2003 and reports correlations of 71 and 53 per cent in different specifications,
while Dobbelaere (2004) reports a correlation of 87 per cent for a set of Belgian firms in the period
1988-1995. The latter article presents two alternative explanations for the positive correlation between
price-cost margins and workers’ bargaining power. One explanation is that a high bargaining power
leads to increased wages and the reduction of the rents kept by the firm. Consequently, some firms
exit the market, thus reducing the intensity of competition in the product market. On the contrary, it
can be argued that workers tend to exert less bargaining pressure if there is no surplus to be extracted
from the firm, which is the case when there is strong competition in the product market. In this context,
Blanchard and Giavazzi (2003) suggest a model that relates labour and product market imperfections.
Chart 6
BARGAINING POWER FOR EACH MARKET IN THE PERIOD 2006-2009
a) Benchmark specification
35
30
b) Alternative specifications
35
Tradable
Non-tradable
Average ( Tradable and non-tradable)
30
25
25
20
20
15
15
10
Per cent
BANCO DE PORTUGAL | ECONOMIC BULLETIN • Spring 2013
84
ficant estimates, i.e., markets where it is not possible to reject the existence of perfect competition in
Per cent
III
OLS
Fixed effects
Random effects
Two-step Heckman
10
5
5
0
0
-5
-5
-10
-10
Source: Author’s calculations.
Notes: Each market corresponds to a 3 digit level in NACE Rev. 1 classification. Black bars identify non-tradable markets as defined
in Amador and Soares (2012a). Coefficients were obtained through OLS regressions with cluster errors, for each market (benchmark
specification). Grey bars correspond to coefficients not significant at 5 per cent, in at least one specification.
Chart 7
PRODUCT AND LABOUR MARKET IMPERFECTION | PER CENT
III
35
30
85
Articles
Bargaining power
25
20
15
10
5
0
Tradable
Non-tradable
-5
-10
0
5
10
15
20
25
30
35
40
45
50
55
60
65
Price-cost margin
Source: Author’s calculations.
Notes: Each market corresponds to a 3 digit level in NACE Rev. 1 classification. Black dots identify non-tradable markets as defined
in Amador and Soares (2012a). Coefficients were obtained through OLS regressions with cluster errors, for each market.
The top block of Table 1 reports estimated price-cost margins, aggregating markets into sectors and
considering several weights (markets, sales, GVA and employment).7 Similarly, the bottom block of the
table displays workers’ bargaining power. “Electricity” and “Construction” exhibit the highest price-cost
margins (above 35 per cent) and are associated to workers’ bargaining power above that of other sectors
of the economy (around 14 and 20 per cent, respectively). In contrast, the lowest price-cost margins are
associated to “Trade” and, to a lesser extent, the manufacturing sector. In these cases, the bargaining
power is also lower than that of other sectors of the Portuguese economy. Furthermore, results obtained
with various weighing variables and alternative specifications are not substantially changed.
Studies for other countries report estimates for price-cost margins and bargaining power. However, the
articles exhibit substantial differences in terms of sectors included, sample periods, characteristics of
the databases and methodological details, which limits comparability. Estrada (2009) uses industry data
and reports price-cost margins for Germany, Spain, Italy and France of 34.7, 25.3, 22.8 and 16.2 per
cent, respectively, and workers’ bargaining power of 20.2, 7.2, 12.6 and 14.2 per cent, respectively.
Additionally, Moreno and Rodriguez (2010) uses a sample of 2000 Spanish manufacturing firms in the
period 1990-2005 and reports a price-cost margin under imperfect labour markets of 17.6 per cent and
a coefficient for workers’ bargaining power that lies between 13 and 15 per cent. Similarly, Dobbelaere
(2004) and Abraham et al. (2009) report an average price-cost margin of 33 to 26 per cent for the Belgian
manufacturing sector, along with a bargaining power of 24 and 12 per cent, respectively. Considering
a set of French firms in the manufacturing sector, Crépon et al. (2005) reports a price-cost margin of 30
per cent and a high parameter for workers’ bargaining power (66 per cent).
7 The weights used are based on the average period of 2006-2009.
Table 1
PRICE-COST MARGIN AND BARGAINING POWER FOR SOME SECTORS
III
Nb. of
markets
(1)
BANCO DE PORTUGAL | ECONOMIC BULLETIN • Spring 2013
86
Non-rejection
of perfect
competition
(percentage of
markets) (2)
Min.
5
6.1
Max.
Median Unweighted
average
Weighted average
Sales
GVA
Employment
26.6
24.9
27.7
25.7
(5.4)
(3.1)
(4.2)
(1.9)
24.2
25.3
24.7
Price-cost margin
Overall economy
156
61.7
25.2
Manufacturing
93
9
6.1
46.8
24.8
24.7
(6.4)
(5.5)
(4.6)
(3.0)
Non-manufacturing
63
0
7.7
61.7
27.8
29.5
25.3
28.8
26.2
(5.4)
(2.8)
(4.1)
(1.6)
Tradable
108
7
6.1
56.1
25.0
25.8
24.7
25.7
25.4
(6.2)
(4.8)
(4.0)
(2.6)
Non-tradable
48
0
7.7
61.7
26.9
28.5
25.1
29.3
25.9
(3.7)
(2.8)
(4.2)
(1.7)
Electricity and water supply
3
0
29.6
39.2
38.6
35.8
38.0
38.1
38.5
(6.6)
(6.6)
(6.6)
(6.7)
Construction
5
0
28.3
47.5
39.3
38.9
44.6
44.1
43.2
(2.8)
(0.7)
(0.7)
(0.7)
Trade
23
0
7.7
57.7
19.0
20.9
17.2
19.7
20.4
(1.8)
(0.9)
(0.9)
(1.0)
31.7
26.8
26.3
27.5
(6.5)
(5.0)
(5.1)
(3.7)
34.4
32.8
30.3
21.8
(3.9)
(1.7)
(1.7)
(1.7)
13.5
11.9
12.9
12.8
(5.2)
(2.6)
(3.4)
(2.2)
11.8
13.0
13.4
Transports and
communications
Other services
10
22
0
0
21.4
9.2
56.1
61.7
27.8
34.0
Bargaining power
Overall economy
156
24
-8.6
34.1
13.5
Manufacturing
93
30
-8.6
30.7
13.8
13.1
(5.8)
(5.6)
(4.4)
(2.9)
Non-manufacturing
63
14
-1.2
34.1
12.3
14.0
11.9
12.8
12.4
(5.2)
(2.2)
(3.3)
(2.0)
Tradable
108
28
-8.6
34.1
13.9
13.5
11.5
11.8
12.7
(5.6)
(5.0)
(4.0)
(2.5)
Non-tradable
48
15
-1.2
27.0
12.2
13.5
12.2
13.7
12.8
(3.7)
(2.1)
(3.3)
(2.1)
Electricity and water supply
3
67
7.6
25.7
8.6
14.0
9.7
10.5
16.0
(6.7)
(4.5)
(4.5)
(4.7)
Construction
5
0
16.0
24.7
19.1
20.6
23.4
23.2
22.8
(2.4)
(0.6)
(0.6)
(0.6)
Trade
23
4
4.7
27.0
10.0
11.4
9.4
10.9
11.6
(1.7)
(0.8)
(0.8)
(1.0)
16.1
13.5
12.7
13.0
(5.0)
(4.4)
(4.5)
(3.2)
14.2
11.6
9.7
6.0
(4.0)
(1.8)
(2.2)
(3.5)
Transports and
communications
Other services
10
22
20
18
5.3
-1.2
34.1
30.3
16.4
14.5
Source: Author’s calculations.
Notes: (1) Each market corresponds to a 3 digit level in NACE Rev. 1. Coefficients were obtained by OLS with cluster errors, for each
market. Standard errors, reported in parenthesis, were computed using the delta method (Greene (1993)). (2) The non-rejection of
the hypothesis of perfect competition is evaluated at a significance level of 5 per cent.
5. Conclusions
This article is based on the methodology proposed by Roeger (1995) to estimate price-cost margins in
the Portuguese economy for the period 2006-2009, allowing for imperfect competition in the labour
III
market. The perfect competition paradigm is widely rejected in the Portuguese economy both in product
The hypothesis of perfect competition in the product market is not rejected in only 5 per cent of the
markets. Estimated price-cost margins are very heterogeneous across markets and figures for the
overall economy range between 25 and 28 per cent, depending on the weight used for each individual
market. In addition, the price-cost margin in the tradable sector is lower than the one observed in the
non-tradable, consistently with the pattern observed in previous studies. Moreover, disregarding labour
market imperfection implies that the price-cost margin is underestimated on average by 11 p.p..
In approximately 25 per cent of the markets, the hypothesis of perfect competition in the labour market
cannot be rejected. The average workers’ bargaining power in the Portuguese economy lies between
12 and 14 per cent, depending on the weight used for each market. Additionally, there is substantial
heterogeneity across sectors, reaching higher values for “Construction” and “Transports and Communications”. Finally, as mentioned in the literature, workers’ bargaining power is strongly and positively
correlated with the price-cost margin across markets.
This article confirms previous findings on the existence of a significant scope to improve competition in
Portuguese product markets, particularly in the non-tradable sector. The inexistence of a suitable competitive setup in the past may have favoured an over allocation of resources in this sector. Therefore, the
improvement of competition is a crucial condition for a successful and sustainable adjustment process
in the Portuguese economy, based on an efficient allocation of resources across firms and markets.
References
Abraham, F., Konings, J. and Vanormelingen, S. (2009), “The effect of globalization on union bargaining
and price-cost margins of firms”, Review of World Economics 145(1), 13–36.
Amador, J. and Soares, A. (2013), “Competition in the Portuguese economy: Estimated price-cost margins under imperfect labour markets”, forthcoming in Banco de Portugal Working Paper series.
Amador, J. and Soares, A. (2012a), “Competition in the Portuguese economy: An overview of classical
indicators”, Working Paper 8, Banco de Portugal.
Amador, J. and Soares, A. (2012b), “Competition in the Portuguese economy: Insights from a profit
elasticity approach”, Working Paper 17, Banco de Portugal.
Bassanetti, A., Torrini, R. and Zollino, F. (2010), “Changing institutions in the european market: the impact on mark-ups and rents allocation”, Working Papers 11, Department of the Treasury, Ministry
of the Economy and of Finance.
Blanchard, O. and Giavazzi, F. (2003), “Macroeconomic effects of regulation and deregulation in goods
and labor markets”, The Quarterly Journal of Economics 118(3), 879–907.
Boulhol, H., Dobbelaere, S. and Maioli, S. (2006), “Imports as product and labour market discipline”,
Discussion Paper 2178, IZA.
Christopoulou, R. and Vermeulen, P. (2012), “Markups in the euro area and the US over the period 19812004: A comparison of 50 sectors”, Empirical Economics 42(1), 53–77.
Crépon, B., Desplatz, R. and Mairesse, J. (2005), “Price-cost margins and rent sharing: Evidence from a
panel of French manufacturing firms”, Annals of Economics and Statistics (79/80), 583–610.
87
Articles
and labour markets.
Dobbelaere (2004), “Estimation of price-cost margins and union bargaining power for Belgian manufacturing”, International Journal of Industrial Organization 22(10), 1381–1398.
III
BANCO DE PORTUGAL | ECONOMIC BULLETIN • Spring 2013
88
Estrada, A. (2009), “The mark-ups in the Spanish economy: International comparison and recent evolution”, Banco de Espãna Working Papers 0905, Banco de Espãna.
Frederick, H. and Harris, B. (1988), “Capital intensity and the firm’s cost of capital”, The Review of Economics and Statistics 70(4), 587–594.
Greene, W. (1993), Econometric Analysis, Collier Macmillan.
Görg, H. and Warzynski, F. (2006), “The dynamics of price cost margins: Evidence from UK manufacturing”, Revue de l’OFCE 97(5), 303–318.
Hall, R. (1988), “The relation between price and marginal cost in US industry”, The Journal of Political
Economy 96(5), 921–947.
Heckman, J. (1979), “Sample selection bias as a specification error”, Econometrica 47(1), 153–161.
Jorgenson, D. and Hall, R. (1967), “Tax policy and investment behaviour”, American Economic Review
57(3), 391–414.
Konings, J. and Vandenbussche, H. (2005), “Antidumping protection and markups of domestic firms”,
Journal of International Economics 65(1), 151–165.
Levinsohn, J. (1993), “Testing the imports-as-market-discipline hypothesis”, Journal of International Economics 35(1-2), 1–22.
Loecker, J. D. and Warzynski, F. (2012), “Markups and firm-level export status”, American Economic
Review 102(6), 2437–71.
Moreno, L. and Rodríguez, D. (2010), “Markups, bargaining power and offshoring: An empirical assessment”, Working Papers 10-05, Asociaciõn Espãnola de Economía y Finanzas Internacionales.
OECD (2001), “Measuring Capital, OECD Manual: Measurement of capital stocks, consumption of fixed
capital and capital services”, OECD.
Olley, G. S. and Pakes, A. (1996), “The dynamics of productivity in the telecommunications equipment
industry”, Econometrica 64(6), 1263–97.
Roeger, W. (1995), “Can imperfect competition explain the difference between primal and dual productivity measures? Estimates for US manufacturing”, The Journal of Political Economy 103(2),
316–330.
Solow, R. (1957), “Technical change and the aggregate production function”, Review of Economics and
Statistics 39(3), 312–320.